Lab 3¶
1, Obtain and download the data¶
The data is downloaded from Kaggle. The dataset contains images of dogs and cats. The data will be divided into training, validation, and test sets. There are totally 5000 images. Each image has a resolution of 180×180. Sources come from: (https://www.kaggle.com/datasets/biaiscience/dogs-vs-cats, with Open Data license)
In [1]:
from google.colab import drive
drive.mount('/content/drive')
Mounted at /content/drive
In [2]:
import zipfile
zip_file_path = "/content/drive/MyDrive/train.zip"
if zipfile.is_zipfile(zip_file_path):
print("valid")
else:
print("not valid")
valid
In [3]:
import zipfile
import os
zip_file_path = "/content/drive/MyDrive/train.zip"
extract_to_path = "/content/train"
if os.path.exists(zip_file_path):
with zipfile.ZipFile(zip_file_path, 'r') as zip_ref:
zip_ref.extractall(extract_to_path)
print(f"zip in {extract_to_path}")
else:
print(f"zip {zip_file_path} dont exist")
zip in /content/train
In [4]:
import os, shutil, pathlib
original_dir = pathlib.Path("/content/train/train")
new_base_dir = pathlib.Path("/content/kaggle_dogs_vs_cats")
def make_subset(subset_name, start_index, end_index):
for category in ("cat", "dog"):
dir = new_base_dir / subset_name / category
os.makedirs(dir)
fnames = [f"{category}.{i}.jpg" for i in range(start_index, end_index)]
for fname in fnames:
shutil.copyfile(src=original_dir / fname,
dst=dir / fname)
make_subset("train", start_index=0, end_index=4000)
make_subset("validation", start_index=4000, end_index=4750)
make_subset("test", start_index=4750, end_index=5000)
2. EDA¶
Basic information and distribution¶
In [5]:
import os
import pandas as pd
from pathlib import Path
import matplotlib.pyplot as plt
import seaborn as sns
new_base_dir = Path("/content/kaggle_dogs_vs_cats")
subsets = ["train", "validation", "test"]
categories = ["cat", "dog"]
data = {"subset": [], "category": [], "file_count": []}
for subset in subsets:
for category in categories:
dir_path = new_base_dir / subset / category
if dir_path.exists():
file_count = len(os.listdir(dir_path))
else:
file_count = 0
data["subset"].append(subset)
data["category"].append(category)
data["file_count"].append(file_count)
df = pd.DataFrame(data)
print("Summary:")
print(df)
plt.figure(figsize=(11, 6))
sns.barplot(x="subset", y="file_count", hue="category", data=df, palette="muted")
plt.title("File Count Distribution Across Subsets and Categories")
plt.ylabel("File Count")
plt.xlabel("Subset")
plt.show()
Summary:
subset category file_count
0 train cat 4000
1 train dog 4000
2 validation cat 750
3 validation dog 750
4 test cat 250
5 test dog 250
Display sample images¶
In [15]:
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import random
from pathlib import Path
def display_sample_images(base_dir, subset, categories, num_images=12):
plt.figure(figsize=(12, 10))
images_per_category = num_images // len(categories)
img_index = 1
for category in categories:
category_dir = base_dir / subset / category
all_images = list(category_dir.glob("*.jpg"))
selected_images = random.sample(all_images, images_per_category)
for img_path in selected_images:
ax = plt.subplot(len(categories), images_per_category, img_index)
img = mpimg.imread(img_path)
plt.imshow(img)
plt.title(f"{category}")
plt.axis("off")
img_index += 1
plt.tight_layout()
plt.show()
# 使用函数显示训练集的样本图片
display_sample_images(
base_dir=new_base_dir,
subset="train",
categories=categories,
num_images=12
)
3. Train two networks¶
SqueezeNet¶
In [9]:
from tensorflow import keras
from tensorflow.keras import layers
from tensorflow.keras.utils import image_dataset_from_directory
import pathlib
data_folder = pathlib.Path("/content/kaggle_dogs_vs_cats")
squeezenet_train_dataset = image_dataset_from_directory(
data_folder / "train",
image_size=(224, 224), # SqueezeNet 224x224
batch_size=32
)
squeezenet_validation_dataset = image_dataset_from_directory(
data_folder / "validation",
image_size=(224, 224),
batch_size=32
)
squeezenet_test_dataset = image_dataset_from_directory(
data_folder / "test",
image_size=(224, 224),
batch_size=32
)
def squeezenet_preprocess(images, labels):
images = images / 255.0 #
return images, labels
squeezenet_train_dataset = squeezenet_train_dataset.map(squeezenet_preprocess)
squeezenet_validation_dataset = squeezenet_validation_dataset.map(squeezenet_preprocess)
squeezenet_test_dataset = squeezenet_test_dataset.map(squeezenet_preprocess)
def squeezenet_fire_module(x, squeeze_channels, expand_channels):
squeeze = layers.Conv2D(squeeze_channels, (1, 1), activation='relu')(x)
expand1x1 = layers.Conv2D(expand_channels, (1, 1), activation='relu')(squeeze)
expand3x3 = layers.Conv2D(expand_channels, (3, 3), padding='same', activation='relu')(squeeze)
return layers.Concatenate()([expand1x1, expand3x3])
def build_squeezenet(input_shape=(224, 224, 3)):
inputs = keras.Input(shape=input_shape)
# Initial Convolution
x = layers.Conv2D(96, (7, 7), strides=(2, 2), activation='relu')(inputs)
x = layers.MaxPooling2D((3, 3), strides=(2, 2))(x)
# Fire Modules
x = squeezenet_fire_module(x, squeeze_channels=16, expand_channels=64)
x = squeezenet_fire_module(x, squeeze_channels=16, expand_channels=64)
x = squeezenet_fire_module(x, squeeze_channels=32, expand_channels=128)
x = layers.MaxPooling2D((3, 3), strides=(2, 2))(x)
x = squeezenet_fire_module(x, squeeze_channels=32, expand_channels=128)
x = squeezenet_fire_module(x, squeeze_channels=48, expand_channels=192)
x = squeezenet_fire_module(x, squeeze_channels=48, expand_channels=192)
x = squeezenet_fire_module(x, squeeze_channels=64, expand_channels=256)
x = layers.MaxPooling2D((3, 3), strides=(2, 2))(x)
# Final Convolution
x = layers.Conv2D(1, (1, 1), activation='relu')(x)
x = layers.GlobalAveragePooling2D()(x)
outputs = layers.Activation('sigmoid')(x)
return keras.Model(inputs, outputs)
squeezenet_model = build_squeezenet()
squeezenet_model.summary()
squeezenet_model.compile(
loss="binary_crossentropy",
optimizer="adam",
metrics=["accuracy"]
)
# recall function
squeezenet_callbacks = [
keras.callbacks.ModelCheckpoint(
filepath="./models/squeezenet.keras",
save_best_only=True,
monitor="val_loss"
)
]
squeezenet_history = squeezenet_model.fit(
squeezenet_train_dataset,
epochs=5,
validation_data=squeezenet_validation_dataset,
callbacks=squeezenet_callbacks
)
squeezenet_test_loss, squeezenet_test_accuracy = squeezenet_model.evaluate(squeezenet_test_dataset)
print(f"SqueezeNet Test Accuracy: {squeezenet_test_accuracy:.2f}")
Found 8000 files belonging to 2 classes. Found 1500 files belonging to 2 classes. Found 500 files belonging to 2 classes.
Model: "functional_2"
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┓ ┃ Layer (type) ┃ Output Shape ┃ Param # ┃ Connected to ┃ ┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━┩ │ input_layer_3 │ (None, 224, 224, 3) │ 0 │ - │ │ (InputLayer) │ │ │ │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_23 (Conv2D) │ (None, 109, 109, 96) │ 14,208 │ input_layer_3[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ max_pooling2d_3 │ (None, 54, 54, 96) │ 0 │ conv2d_23[0][0] │ │ (MaxPooling2D) │ │ │ │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_24 (Conv2D) │ (None, 54, 54, 16) │ 1,552 │ max_pooling2d_3[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_25 (Conv2D) │ (None, 54, 54, 64) │ 1,088 │ conv2d_24[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_26 (Conv2D) │ (None, 54, 54, 64) │ 9,280 │ conv2d_24[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ concatenate_7 │ (None, 54, 54, 128) │ 0 │ conv2d_25[0][0], │ │ (Concatenate) │ │ │ conv2d_26[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_27 (Conv2D) │ (None, 54, 54, 16) │ 2,064 │ concatenate_7[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_28 (Conv2D) │ (None, 54, 54, 64) │ 1,088 │ conv2d_27[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_29 (Conv2D) │ (None, 54, 54, 64) │ 9,280 │ conv2d_27[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ concatenate_8 │ (None, 54, 54, 128) │ 0 │ conv2d_28[0][0], │ │ (Concatenate) │ │ │ conv2d_29[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_30 (Conv2D) │ (None, 54, 54, 32) │ 4,128 │ concatenate_8[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_31 (Conv2D) │ (None, 54, 54, 128) │ 4,224 │ conv2d_30[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_32 (Conv2D) │ (None, 54, 54, 128) │ 36,992 │ conv2d_30[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ concatenate_9 │ (None, 54, 54, 256) │ 0 │ conv2d_31[0][0], │ │ (Concatenate) │ │ │ conv2d_32[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ max_pooling2d_4 │ (None, 26, 26, 256) │ 0 │ concatenate_9[0][0] │ │ (MaxPooling2D) │ │ │ │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_33 (Conv2D) │ (None, 26, 26, 32) │ 8,224 │ max_pooling2d_4[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_34 (Conv2D) │ (None, 26, 26, 128) │ 4,224 │ conv2d_33[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_35 (Conv2D) │ (None, 26, 26, 128) │ 36,992 │ conv2d_33[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ concatenate_10 │ (None, 26, 26, 256) │ 0 │ conv2d_34[0][0], │ │ (Concatenate) │ │ │ conv2d_35[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_36 (Conv2D) │ (None, 26, 26, 48) │ 12,336 │ concatenate_10[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_37 (Conv2D) │ (None, 26, 26, 192) │ 9,408 │ conv2d_36[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_38 (Conv2D) │ (None, 26, 26, 192) │ 83,136 │ conv2d_36[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ concatenate_11 │ (None, 26, 26, 384) │ 0 │ conv2d_37[0][0], │ │ (Concatenate) │ │ │ conv2d_38[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_39 (Conv2D) │ (None, 26, 26, 48) │ 18,480 │ concatenate_11[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_40 (Conv2D) │ (None, 26, 26, 192) │ 9,408 │ conv2d_39[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_41 (Conv2D) │ (None, 26, 26, 192) │ 83,136 │ conv2d_39[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ concatenate_12 │ (None, 26, 26, 384) │ 0 │ conv2d_40[0][0], │ │ (Concatenate) │ │ │ conv2d_41[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_42 (Conv2D) │ (None, 26, 26, 64) │ 24,640 │ concatenate_12[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_43 (Conv2D) │ (None, 26, 26, 256) │ 16,640 │ conv2d_42[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_44 (Conv2D) │ (None, 26, 26, 256) │ 147,712 │ conv2d_42[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ concatenate_13 │ (None, 26, 26, 512) │ 0 │ conv2d_43[0][0], │ │ (Concatenate) │ │ │ conv2d_44[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ max_pooling2d_5 │ (None, 12, 12, 512) │ 0 │ concatenate_13[0][0] │ │ (MaxPooling2D) │ │ │ │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ conv2d_45 (Conv2D) │ (None, 12, 12, 1) │ 513 │ max_pooling2d_5[0][0] │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ global_average_pooling2d… │ (None, 1) │ 0 │ conv2d_45[0][0] │ │ (GlobalAveragePooling2D) │ │ │ │ ├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤ │ activation_1 (Activation) │ (None, 1) │ 0 │ global_average_poolin… │ └───────────────────────────┴────────────────────────┴────────────────┴────────────────────────┘
Total params: 538,753 (2.06 MB)
Trainable params: 538,753 (2.06 MB)
Non-trainable params: 0 (0.00 B)
Epoch 1/5 250/250 ━━━━━━━━━━━━━━━━━━━━ 1337s 5s/step - accuracy: 0.4891 - loss: 0.6931 - val_accuracy: 0.5000 - val_loss: 0.6931 Epoch 2/5 250/250 ━━━━━━━━━━━━━━━━━━━━ 1315s 5s/step - accuracy: 0.4876 - loss: 0.6931 - val_accuracy: 0.5000 - val_loss: 0.6931 Epoch 3/5 250/250 ━━━━━━━━━━━━━━━━━━━━ 1338s 5s/step - accuracy: 0.4896 - loss: 0.6931 - val_accuracy: 0.5000 - val_loss: 0.6931 Epoch 4/5 250/250 ━━━━━━━━━━━━━━━━━━━━ 1333s 5s/step - accuracy: 0.4873 - loss: 0.6931 - val_accuracy: 0.5000 - val_loss: 0.6931 Epoch 5/5 250/250 ━━━━━━━━━━━━━━━━━━━━ 1298s 5s/step - accuracy: 0.4879 - loss: 0.6931 - val_accuracy: 0.5000 - val_loss: 0.6931 16/16 ━━━━━━━━━━━━━━━━━━━━ 27s 2s/step - accuracy: 0.5247 - loss: 0.6931 SqueezeNet Test Accuracy: 0.50
Training and Validation Accuracy & Loss
In [17]:
import matplotlib.pyplot as plt
def plot_history(history):
accuracy = history.history["accuracy"]
val_accuracy = history.history["val_accuracy"]
loss = history.history["loss"]
val_loss = history.history["val_loss"]
epochs_range = range(1, len(accuracy) + 1)
plt.figure(figsize=(16, 6))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, accuracy, "bo", label="Training accuracy")
plt.plot(epochs_range, val_accuracy, "b", label="Validation accuracy")
plt.xlabel("Epochs")
plt.ylabel("Accuracy")
plt.title("Training and Validation Accuracy")
plt.legend(loc="upper left")
plt.subplot(1, 2, 2)
plt.plot(epochs_range, loss, "bo", label="Training loss")
plt.plot(epochs_range, val_loss, "b", label="Validation loss")
plt.xlabel("Epochs")
plt.ylabel("Loss")
plt.title("Training and Validation Loss")
plt.legend(loc="upper left")
plt.show()
plot_history(squeezenet_history)
Vgg 16¶
In [7]:
from tensorflow import keras
from tensorflow.keras import layers
import pathlib
import numpy as np
from tensorflow.keras.utils import image_dataset_from_directory
data_folder = pathlib.Path("/content/kaggle_dogs_vs_cats")
train_dataset = image_dataset_from_directory(
data_folder / "train",
image_size=(180, 180),
batch_size=32
)
validation_dataset = image_dataset_from_directory(
data_folder / "validation",
image_size=(180, 180),
batch_size=32
)
test_dataset = image_dataset_from_directory(
data_folder / "test",
image_size=(180, 180),
batch_size=32
)
conv_base = keras.applications.VGG16(
weights="imagenet",
include_top=False,
input_shape=(180, 180, 3)
)
conv_base.summary()
def get_features_and_labels(dataset):
all_features = []
all_labels = []
for images, labels in dataset:
preprocessed_images = keras.applications.vgg16.preprocess_input(images)
features = conv_base.predict(preprocessed_images)
all_features.append(features)
all_labels.append(labels)
return np.concatenate(all_features), np.concatenate(all_labels)
train_features, train_labels = get_features_and_labels(train_dataset)
val_features, val_labels = get_features_and_labels(validation_dataset)
test_features, test_labels = get_features_and_labels(test_dataset)
print("Train features shape:", train_features.shape)
inputs = keras.Input(shape=(5, 5, 512))
x = layers.Flatten()(inputs)
x = layers.Dense(256, activation="relu")(x)
x = layers.Dropout(0.5)(x)
outputs = layers.Dense(1, activation="sigmoid")(x)
model = keras.Model(inputs, outputs)
model.summary()
model.compile(
loss="binary_crossentropy",
optimizer="rmsprop",
metrics=["accuracy"]
)
callbacks = [
keras.callbacks.ModelCheckpoint(
filepath="./models/vgg16_fine_tuned.keras",
save_best_only=True,
monitor="val_loss"
)
]
history = model.fit(
train_features,
train_labels,
epochs=30,
validation_data=(val_features, val_labels),
callbacks=callbacks
)
test_loss, test_accuracy = model.evaluate(test_features, test_labels)
print(f"Test Accuracy: {test_accuracy:.2f}")
Found 8000 files belonging to 2 classes. Found 1500 files belonging to 2 classes. Found 500 files belonging to 2 classes. Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5 58889256/58889256 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
Model: "vgg16"
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓ ┃ Layer (type) ┃ Output Shape ┃ Param # ┃ ┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩ │ input_layer_1 (InputLayer) │ (None, 180, 180, 3) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block1_conv1 (Conv2D) │ (None, 180, 180, 64) │ 1,792 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block1_conv2 (Conv2D) │ (None, 180, 180, 64) │ 36,928 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block1_pool (MaxPooling2D) │ (None, 90, 90, 64) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block2_conv1 (Conv2D) │ (None, 90, 90, 128) │ 73,856 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block2_conv2 (Conv2D) │ (None, 90, 90, 128) │ 147,584 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block2_pool (MaxPooling2D) │ (None, 45, 45, 128) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block3_conv1 (Conv2D) │ (None, 45, 45, 256) │ 295,168 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block3_conv2 (Conv2D) │ (None, 45, 45, 256) │ 590,080 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block3_conv3 (Conv2D) │ (None, 45, 45, 256) │ 590,080 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block3_pool (MaxPooling2D) │ (None, 22, 22, 256) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block4_conv1 (Conv2D) │ (None, 22, 22, 512) │ 1,180,160 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block4_conv2 (Conv2D) │ (None, 22, 22, 512) │ 2,359,808 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block4_conv3 (Conv2D) │ (None, 22, 22, 512) │ 2,359,808 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block4_pool (MaxPooling2D) │ (None, 11, 11, 512) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block5_conv1 (Conv2D) │ (None, 11, 11, 512) │ 2,359,808 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block5_conv2 (Conv2D) │ (None, 11, 11, 512) │ 2,359,808 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block5_conv3 (Conv2D) │ (None, 11, 11, 512) │ 2,359,808 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ block5_pool (MaxPooling2D) │ (None, 5, 5, 512) │ 0 │ └──────────────────────────────────────┴─────────────────────────────┴─────────────────┘
Total params: 14,714,688 (56.13 MB)
Trainable params: 14,714,688 (56.13 MB)
Non-trainable params: 0 (0.00 B)
1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 15s 15s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 15s 15s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 15s 15s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 15s 15s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 15s 15s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 15s 15s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 9s 9s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 14s 14s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 6s 6s/step Train features shape: (8000, 5, 5, 512)
Model: "functional_1"
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓ ┃ Layer (type) ┃ Output Shape ┃ Param # ┃ ┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩ │ input_layer_2 (InputLayer) │ (None, 5, 5, 512) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ flatten (Flatten) │ (None, 12800) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ dense (Dense) │ (None, 256) │ 3,277,056 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ dropout (Dropout) │ (None, 256) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ dense_1 (Dense) │ (None, 1) │ 257 │ └──────────────────────────────────────┴─────────────────────────────┴─────────────────┘
Total params: 3,277,313 (12.50 MB)
Trainable params: 3,277,313 (12.50 MB)
Non-trainable params: 0 (0.00 B)
Epoch 1/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 11s 40ms/step - accuracy: 0.9162 - loss: 4.0252 - val_accuracy: 0.9693 - val_loss: 0.5029 Epoch 2/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 11s 43ms/step - accuracy: 0.9757 - loss: 0.2367 - val_accuracy: 0.9780 - val_loss: 0.2647 Epoch 3/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 19s 37ms/step - accuracy: 0.9826 - loss: 0.1578 - val_accuracy: 0.9767 - val_loss: 0.3452 Epoch 4/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 12s 45ms/step - accuracy: 0.9929 - loss: 0.0579 - val_accuracy: 0.9807 - val_loss: 0.3436 Epoch 5/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 21s 48ms/step - accuracy: 0.9914 - loss: 0.0670 - val_accuracy: 0.9793 - val_loss: 0.3576 Epoch 6/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 19s 43ms/step - accuracy: 0.9918 - loss: 0.0671 - val_accuracy: 0.9773 - val_loss: 0.4130 Epoch 7/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 22s 51ms/step - accuracy: 0.9934 - loss: 0.0470 - val_accuracy: 0.9780 - val_loss: 0.3775 Epoch 8/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 19s 46ms/step - accuracy: 0.9950 - loss: 0.0390 - val_accuracy: 0.9813 - val_loss: 0.3821 Epoch 9/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 11s 43ms/step - accuracy: 0.9972 - loss: 0.0306 - val_accuracy: 0.9780 - val_loss: 0.4135 Epoch 10/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 24s 58ms/step - accuracy: 0.9971 - loss: 0.0212 - val_accuracy: 0.9753 - val_loss: 0.4291 Epoch 11/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 18s 48ms/step - accuracy: 0.9953 - loss: 0.0446 - val_accuracy: 0.9787 - val_loss: 0.5136 Epoch 12/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 19s 42ms/step - accuracy: 0.9962 - loss: 0.0327 - val_accuracy: 0.9813 - val_loss: 0.4999 Epoch 13/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 12s 47ms/step - accuracy: 0.9967 - loss: 0.0304 - val_accuracy: 0.9753 - val_loss: 0.5704 Epoch 14/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 20s 47ms/step - accuracy: 0.9980 - loss: 0.0165 - val_accuracy: 0.9767 - val_loss: 0.4956 Epoch 15/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 21s 48ms/step - accuracy: 0.9973 - loss: 0.0197 - val_accuracy: 0.9727 - val_loss: 0.5516 Epoch 16/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 12s 50ms/step - accuracy: 0.9982 - loss: 0.0052 - val_accuracy: 0.9767 - val_loss: 0.4883 Epoch 17/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 11s 42ms/step - accuracy: 0.9992 - loss: 0.0036 - val_accuracy: 0.9780 - val_loss: 0.5117 Epoch 18/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 23s 51ms/step - accuracy: 0.9969 - loss: 0.0401 - val_accuracy: 0.9787 - val_loss: 0.5626 Epoch 19/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 19s 46ms/step - accuracy: 0.9985 - loss: 0.0157 - val_accuracy: 0.9787 - val_loss: 0.5821 Epoch 20/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 21s 49ms/step - accuracy: 0.9994 - loss: 0.0021 - val_accuracy: 0.9780 - val_loss: 0.5981 Epoch 21/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 14s 55ms/step - accuracy: 0.9990 - loss: 0.0054 - val_accuracy: 0.9793 - val_loss: 0.5991 Epoch 22/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 19s 50ms/step - accuracy: 0.9998 - loss: 0.0029 - val_accuracy: 0.9767 - val_loss: 0.5699 Epoch 23/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 20s 49ms/step - accuracy: 0.9993 - loss: 0.0044 - val_accuracy: 0.9787 - val_loss: 0.5775 Epoch 24/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 20s 47ms/step - accuracy: 0.9983 - loss: 0.0197 - val_accuracy: 0.9800 - val_loss: 0.6149 Epoch 25/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 20s 47ms/step - accuracy: 0.9998 - loss: 0.0022 - val_accuracy: 0.9800 - val_loss: 0.5827 Epoch 26/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 12s 48ms/step - accuracy: 0.9999 - loss: 2.6011e-04 - val_accuracy: 0.9813 - val_loss: 0.5648 Epoch 27/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 21s 49ms/step - accuracy: 0.9990 - loss: 0.0040 - val_accuracy: 0.9813 - val_loss: 0.5296 Epoch 28/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 19s 44ms/step - accuracy: 0.9993 - loss: 0.0070 - val_accuracy: 0.9793 - val_loss: 0.5627 Epoch 29/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 12s 50ms/step - accuracy: 0.9994 - loss: 0.0013 - val_accuracy: 0.9800 - val_loss: 0.5910 Epoch 30/30 250/250 ━━━━━━━━━━━━━━━━━━━━ 12s 47ms/step - accuracy: 0.9990 - loss: 0.0082 - val_accuracy: 0.9813 - val_loss: 0.6512 16/16 ━━━━━━━━━━━━━━━━━━━━ 0s 9ms/step - accuracy: 0.9744 - loss: 0.2751 Test Accuracy: 0.98
Training and Validation Accuracy & Loss
In [18]:
import matplotlib.pyplot as plt
def plot_vgg16_history(history):
accuracy = history.history["accuracy"]
val_accuracy = history.history["val_accuracy"]
loss = history.history["loss"]
val_loss = history.history["val_loss"]
epochs_range = range(1, len(accuracy) + 1)
plt.figure(figsize=(16, 6))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, accuracy, "bo", label="Training Accuracy")
plt.plot(epochs_range, val_accuracy, "b", label="Validation Accuracy")
plt.xlabel("Epochs")
plt.ylabel("Accuracy")
plt.title("Training and Validation Accuracy")
plt.legend(loc="upper left")
plt.subplot(1, 2, 2)
plt.plot(epochs_range, loss, "bo", label="Training Loss")
plt.plot(epochs_range, val_loss, "b", label="Validation Loss")
plt.xlabel("Epochs")
plt.ylabel("Loss")
plt.title("Training and Validation Loss")
plt.legend(loc="upper left")
plt.show()
plot_vgg16_history(history)
In [20]:
import numpy as np
import matplotlib.pyplot as plt
from sklearn.metrics import confusion_matrix, classification_report, precision_recall_curve
def evaluate_squeezenet(model, dataset, dataset_name):
y_true = []
y_pred = []
for images, labels in dataset:
predictions = model.predict(images)
y_true.extend(labels.numpy())
y_pred.extend((predictions > 0.5).astype(int).flatten())
y_true = np.array(y_true)
y_pred = np.array(y_pred)
accuracy = np.mean(y_true == y_pred)
print(f"{dataset_name} Accuracy: {accuracy:.2f}")
cm = confusion_matrix(y_true, y_pred)
print(f"Confusion Matrix ({dataset_name}):\n", cm)
report = classification_report(y_true, y_pred, target_names=["Cat", "Dog"])
print(f"Classification Report ({dataset_name}):\n", report)
y_proba = model.predict(dataset.map(lambda x, y: x))
precision, recall, _ = precision_recall_curve(y_true, y_proba.flatten())
plt.figure()
plt.plot(recall, precision, label=f"PR Curve ({dataset_name})")
plt.xlabel("Recall")
plt.ylabel("Precision")
plt.title(f"Precision-Recall Curve ({dataset_name})")
plt.legend()
plt.show()
wrong_predictions = np.where(y_true != y_pred)[0]
return y_true, y_pred, wrong_predictions
def analyze_failures_squeezenet(dataset, y_true, y_pred, wrong_indices):
print(f"Analyzing Failures for SqueezeNet...")
for idx in wrong_indices[:5]:
for batch_idx, (images, labels) in enumerate(dataset):
if idx < len(labels):
image = images[idx].numpy()
if image.max() <= 1.0:
image = (image * 255).astype("uint8")
plt.figure()
if image.shape[-1] == 3:
plt.imshow(image)
else:
plt.imshow(image.squeeze(), cmap='gray')
plt.title(f"True: {y_true[idx]} | Predicted: {y_pred[idx]}")
plt.axis('off')
plt.show()
break
idx -= len(labels)
print("Evaluating SqueezeNet...")
y_true_sn, y_pred_sn, wrong_sn = evaluate_squeezenet(squeezenet_model, squeezenet_test_dataset, "SqueezeNet")
analyze_failures_squeezenet(squeezenet_test_dataset, y_true_sn, y_pred_sn, wrong_sn)
Evaluating SqueezeNet... 1/1 ━━━━━━━━━━━━━━━━━━━━ 4s 4s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 4s 4s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 4s 4s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 2s 2s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 2s 2s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 687ms/step SqueezeNet Accuracy: 0.50 Confusion Matrix (SqueezeNet): [[250 0] [250 0]] Classification Report (SqueezeNet): precision recall f1-score support Cat 0.50 1.00 0.67 250 Dog 0.00 0.00 0.00 250 accuracy 0.50 500 macro avg 0.25 0.50 0.33 500 weighted avg 0.25 0.50 0.33 500
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1531: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is", len(result))
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1531: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is", len(result))
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1531: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is", len(result))
16/16 ━━━━━━━━━━━━━━━━━━━━ 21s 1s/step
Analyzing Failures for SqueezeNet...
In [19]:
def evaluate_vgg16(model, dataset, conv_base, dataset_name):
features = []
labels = []
for images, lbls in dataset:
preprocessed_images = keras.applications.vgg16.preprocess_input(images)
feature_batch = conv_base.predict(preprocessed_images)
features.append(feature_batch)
labels.extend(lbls.numpy())
features = np.concatenate(features)
labels = np.array(labels)
predictions = model.predict(features)
y_pred = (predictions > 0.5).astype(int).flatten()
y_true = labels
accuracy = np.mean(y_true == y_pred)
print(f"{dataset_name} Accuracy: {accuracy:.2f}")
cm = confusion_matrix(y_true, y_pred)
print(f"Confusion Matrix ({dataset_name}):\n", cm)
report = classification_report(y_true, y_pred, target_names=["Cat", "Dog"])
print(f"Classification Report ({dataset_name}):\n", report)
precision, recall, _ = precision_recall_curve(y_true, predictions.flatten())
plt.figure()
plt.plot(recall, precision, label=f"PR Curve ({dataset_name})")
plt.xlabel("Recall")
plt.ylabel("Precision")
plt.title(f"Precision-Recall Curve ({dataset_name})")
plt.legend()
plt.show()
wrong_predictions = np.where(y_true != y_pred)[0]
return y_true, y_pred, wrong_predictions
print("Evaluating VGG16...")
y_true_vgg, y_pred_vgg, wrong_vgg = evaluate_vgg16(model, test_dataset, conv_base, "VGG16")
def analyze_failures_vgg16(dataset, conv_base, y_true, y_pred, wrong_indices):
print(f"Analyzing Failures for VGG16...")
for idx in wrong_indices[:5]:
for batch_idx, (images, labels) in enumerate(dataset):
if idx < len(labels):
preprocessed_images = keras.applications.vgg16.preprocess_input(images)
feature_batch = conv_base.predict(preprocessed_images)
plt.figure()
plt.imshow(images[idx].numpy().astype("uint8"))
plt.title(f"True: {y_true[idx]} | Predicted: {y_pred[idx]}")
plt.show()
break
idx -= len(labels)
analyze_failures_vgg16(test_dataset, conv_base, y_true_vgg, y_pred_vgg, wrong_vgg)
Evaluating VGG16... 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 10s 10s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 6s 6s/step 16/16 ━━━━━━━━━━━━━━━━━━━━ 0s 9ms/step VGG16 Accuracy: 0.98 Confusion Matrix (VGG16): [[247 3] [ 6 244]] Classification Report (VGG16): precision recall f1-score support Cat 0.98 0.99 0.98 250 Dog 0.99 0.98 0.98 250 accuracy 0.98 500 macro avg 0.98 0.98 0.98 500 weighted avg 0.98 0.98 0.98 500
Analyzing Failures for VGG16... 1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 12s 12s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 13s 13s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step
5. Conclusions¶
SqeezeNet¶
The SqueezeNet model trained on the dataset of classify dogs and cats images achieved an overall accuracy of 50%. It is equivalent to random guessing. The precision-recall curve seem don't have any change. The curve also illustrates the model's inability to distinguish between the two classes. This could due to the model's poor performance stems from data issues, SqueezeNet's limitations, and training inefficiencies.
While chacking the Confusion Matrix, the model classified all "Cat" images correctly but failed entirely to classify "Dog" images. This might show that a significant class imbalance in the predictions, though not necessarily in the dataset. The other result testinga all show the same thing.
Vgg16¶
The VGG16 model achieved outstanding performance. There is an overall accuracy of 98%. The confusion matrix and classification report indicate balanced performance across both classes. Also while checking with precision, recall, and F1-scores close to 0.98 for both cats and dogs. The PR curve further demonstrates the model's excellent ability to distinguish between the two classes.
Vgg16 has a stronger ability on classify the images on this dataset. The conclusion can also be the model's complexity is well-suited to capture the nuanced differences between the two classes.